Re: consistant failures on a new kind of spam

FYI, I do effectively the same thing with "ASSP" 
http://assp.sourceforge.net/

I have ASSP setup server-wide, which most users prefer (and I prefer *for 
them* due to its single system config),
however it has precisely the same downside as you describe for CRM ... 
particularly annoying to me is that its "too aggressive" and tends to mark 
emails as spam too often, so I use the mode where it simply marks, but not 
discards, spam and leave it to the delivery agent (procmail) to filter into 
spam boxes.

Then I Cross-train with spamprobe ... I like the result ...
Practically nothing makes it through both filters ... (maybe  1 of 1000 ? ) 
although some ham is marked spam by ASSP so I do wad through maybe 20 
emails/day in the mailboxes where I store spam which was only caught by one 
of the 2 filters.   I then move the falsely categorized ham or spam and run 
a script which trains both ASSP and spamprobe on their respective mistakes.





> I've been using CRM-114 for about 5 months to catch all the spams that
> Spamprobe is missing.  I continually feedback all the missed spams (from
> both sides) into both Spamprobe and CRM-114 for training.  Spamprobe
> still manages to miss 30-50 spams a day which CRM-114 is able to catch.
>
> The downside:
> o CRM-114 has more false-positives than Spamprobe.  But it does seem to
> be learning.
> o CRM-114 is a pain to set up and get running.
> o CRM-114 misses a lot of spam everyday (classifies it as "unsure").
> But the Unsure mails are mostly spam and easy to wade through.
>
> Phil
>
>
> Brian Burton wrote:
>> Darxus-5+gNnNaL/BD/[email protected] wrote:
>>
>>> On 08/03, Darxus-5+gNnNaL/BD/[email protected] wrote:
>>>
>>>> I've started getting spams that I can easily identify by appearance,
>>>> and spamprobe has flagged them *all* as good, even though I've 
>>>> retrained
>>>> it on each one as it has come in.
>>>>
>>> I haven't found a solution to this with spamprobe, but the author of 
>>> CRM114
>>> says it always catches these.  My *guess* is that the primary difference 
>>> is
>>> in the tokenizing.
>>>
>>
>> If you give CRM113 a try please let me know how well it works in real
>> life as opposed to what the author reports.
>>
>> Most likely you are describing an image spam with technical jargon as
>> text.  I think SP's image tokenization needs improvement to catch these.
>>
>> Thanks,
>> ++Brian
>>
>>
>> -------------------------------------------------------------------------
>> Using Tomcat but need to do more? Need to support web services, security?
>> Get stuff done quickly with pre-integrated technology to make your job 
>> easier
>> Download IBM WebSphere Application Server v.1.0.1 based on Apache 
>> Geronimo
>> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
>> _______________________________________________
>> Spamprobe-users mailing list
>> Spamprobe-users-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
>> https://lists.sourceforge.net/lists/listinfo/spamprobe-users
>>
>
> -------------------------------------------------------------------------
> Using Tomcat but need to do more? Need to support web services, security?
> Get stuff done quickly with pre-integrated technology to make your job 
> easier
> Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
> _______________________________________________
> Spamprobe-users mailing list
> Spamprobe-users-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
> https://lists.sourceforge.net/lists/listinfo/spamprobe-users
> 


-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642